To improve planning of maxillofacial surgery, a novel method for maxillary, mandibular, and facial-nerve 3-D image reconstruction was implemented and optimized. A set of images acquired by a computed tomography (CT) scanner was segmented to reconstruct the 3-D model of the maxilla and mandible. Particular attention was given to the segmentation of the facial nerve, which was obtained through the gradient-vector-flow (GVF) snake method. After segmentation, precise anatomical 3-D plastic models were fabricated through stereolithography from CT scans of five clinical cases that will undergo either dental implant surgery or bone distraction. For all cases, 3-D models delivered essential visual and tactile information for the planning and simulation of surgery as well as for customized implant preparation.